bmj · 2016-10-02 · unfortunately, reports of cardiovascular (cv) adverse reactions began to...

Confidential: For Review O

nly

Non-steroidal anti-inflammatory drugs and the risk of heart

failure: a nested case-control study from four European countries in the SOS Project

Journal: BMJ

Manuscript ID BMJ.2015.029885

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the Author: 14-Oct-2015

Complete List of Authors: Arfe, Andrea; University of Milano-Bicocca, Department of Statistics and Quantitative Methods Varas-Lorenzo, Cristina; RTI Health Solutions, Epidemiology Nicotra, Federica; University of Milano-Bicocca, Department of Statistics and Quantitative Methods Scotti, Lorenza; University of Milano-Bicocca, Department of Statistics and Quantitative Methods Zambon, Antonella; University of Milano-Bicocca, Department of Statistics and Quantitative Methods Kollhorst, Bianca; Leibniz Institute of Prevention Research and Epidemiology, Clinical Epidemiology Schink, Tania; Leibniz Institute of Prevention Research and Epidemiology,

Clinical Epidemiology Garbe, Edeltraut; Leibniz Institute of Prevention Research and Epidemiology, Clinical Epidemiology Herings, Ron; PHARMO Institute, Straatman, Huub; PHARMO Institute Schade, Rene; Erasmus University Medical Center, Villa, Marco; Local Health Authority ASL Cremona Lucchi, Silvia; Local Health Authority ASL Cremona Valkhoff, Vera; Erasmus MC, Department of Gastroenterology and Hepatology Romio, Silvana; Erasmus University Medical Center, Department of Medical Informatics

Thiessard, Frantz ; University of Bordeaux Segalen Pariente, Antoine; Univ. de Bordeaux, U657; INSERM, U657 Sturkenboom, Miriam; Erasmus University Medical Center, Medical Informatics Corrao, Giovanni; University of Milano-Bicocca, Department of Statistics and Quantitative Methods

Keywords: NSAIDs, COX-1, COX-2, Heart Failure, Nested case-control study, Database

https://mc.manuscriptcentral.com/bmj

BMJ


nly

Page 1 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

1

Non-steroidal anti-inflammatory drugs and the risk of heart failure: a nested

case-control study from four European countries in the SOS Project

Andrea Arfè1 ([email protected]), Cristina Varas Lorenzo2 ([email protected]), Federica Nicotra1

([email protected]), Lorenza Scotti1 ([email protected]), Antonella Zambon

1

([email protected]), Bianca Kollhorst3 ([email protected]), Tania Schink3

([email protected]), Edeltraut Garbe3 ([email protected]), Ron Herings4

([email protected]), Huub Straatman4

([email protected]), René Schade5

([email protected]), Marco Villa6 ([email protected]), Silvia Lucchi6

([email protected]), Vera Valkhoff5 ([email protected]), Silvana Romio5

([email protected]), Frantz Thiessard7 ([email protected]), Martijn Schuemie

5

([email protected]), Antoine Pariente7 ([email protected]), Miriam

Sturkenboom5 ([email protected]), and Giovanni Corrao

1 ([email protected]); on

behalf of the Safety of Non-steroidal Anti-inflammatory Drugs (SOS) project consortium

1 Unit of Biostatistics, Epidemiology and Public Health. University of Milano-Bicocca, Milan, Italy. 2 RTI

Health Solutions, Barcellona, Spain. 3

Leibniz Institute of Prevention Research and Epidemiology, Bremen,

Germany. 4 PHARMO Institute, Utrecht, the Netherlands.

5 Department of Medical Informatics, Erasmus

University Medical Center, Rotterdam, the Netherlands. 6 Local Health Authority ASL Cremona, Cremona,

Italy. 7 University of Bordeaux Segalen, Bordeaux, France.

Key words: NSAIDs, COX-1, COX-2, Heart Failure, Nested case-control study, Database.

Word count: 3,882

Corresponding Author: Giovanni Corrao. Unit of Biostatistics, Epidemiology and Public Health,

Department of Statistics and Quantitative Methods, University of Milano-Bicocca. Via Bicocca

degli Arcimboldi, 8, Edificio U7, 20126 Milano, Italy. Tel.: +39-02-6448-5854. E-Mail:

[email protected]

Page 2 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

2

Abstract

Objectives: To estimate the risk of hospitalization for heart failure (HF) associated with use of

individual Non-Steroidal Anti-Inflammatory Drugs (NSAIDs).

Methods: This is a case-control study, nested in a cohort of adult (≥18 years) new users of

NSAIDs, performed in 5 population-based healthcare databases from 4 European countries (the

Netherlands, Italy, Germany, and the United Kingdom). Overall, 92,163 hospitalizations for HF

were identified and compared with 8,246,403 controls with respect to use of 27 individual NSAIDs,

including 23 traditional NSAIDs (tNSAIDs) and 4 selective COX-2 inhibitors (COXIBs). The

relation between the used dose of selected individual NSAIDs and HF hospitalization risk was also

assessed.

Results: The estimated adjusted odds ratio (OR) of HF hospitalization associated with current use

of any NSAID was 1.20 (95% confidence interval, CI: 1.17 to 1.22). The risk of HF hospitalization

was increased for seven tNSAIDs (diclofenac, ibuprofen, indomethacin, ketorolac, naproxen,

nimesulide and piroxicam) and two COXIBs (etoricoxib and rofecoxib), with ORs (95% CIs)

ranging from 1.16 (1.07, 1.27) for naproxen to 1.83 (1.66, 2.02) for ketorolac. The risk of

hospitalization for HF was doubled by each of these NSAIDs when currently used at very-high

doses. For indomethacin and etoricoxib, even more moderate doses were associated with increased

risk. There was no evidence that celecoxib use increased the risk of HF hospitalization.

Conclusions: The risk of being hospitalized for HF associated with current use of NSAIDs appears

to vary between individual NSAIDs and this effect is dose-dependent. The risk of hospitalization

for HF with a large number of individual NSAID use estimated in this study may help to inform

both clinicians and health regulators.

Page 3 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

3

Introduction

Non-Steroidal Anti-Inflammatory Drugs (NSAIDs) are a broad class of drugs with analgesic and

anti-inflammatory actions exerted through inhibition of the two recognized forms of prostaglandin

G/H synthase (also referred to as cyclo-oxygenase [COX]), namely COX-1 and COX-2 [1].

Because the therapeutic action of these drugs is mostly mediated by inhibition of COX-2, while

their gastrointestinal adverse reactions are largely due to COX-1 inhibition, NSAIDs selectively

inhibiting COX-2 (collectively known as COXIBs) were developed in the 1990s to reduce the risk

of gastrointestinal toxicity [2].

Unfortunately, reports of cardiovascular (CV) adverse reactions began to emerge during 2001-2003

[3,4], and subsequent placebo-controlled trials showed that COXIBs were associated with an

increased risk of atherothrombotic vascular events [5,6]. More recently, though, meta-analyses of

randomized trials and observational studies have shown that the higher CV risk is not restricted to

COXIBs, but also applies to some traditional NSAIDs (tNSAIDs) [7-12].

In addition, although prostaglandins have both vasodilator and vasoconstrictor effects, the overall

effects of the NSAID-mediated inhibition of the prostaglandin synthesis are to increase peripheral

systemic resistance and reduce renal perfusion, glomerular filtration rate, and sodium excretion in

susceptible individuals [13,14]. Taken together, these mechanisms may potentially trigger clinical

HF manifestations, especially in susceptible patients [15].

Compatibly, NSAIDs use was found to be associated with an increased HF risk several randomized

clinical trials [11] and observational studies [16,17]. A large meta-analysis of over 600 randomized

trials showed that COXIBs or high-doses of diclofenac, ibuprofen, or naproxen increased the risk of

HF hospitalization from 1.9-fold to 2.5-fold in comparison with placebo [11]. In light of this

worrying evidence, current guidelines limit the use of NSAIDs in patients predisposed to HF, with a

full contraindication for patients with diagnosed HF [18].

Page 4 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

4

Nonetheless, there is still limited information on the risk of HF associated with the use of individual

NSAIDs (both COXIBs and tNSAIDs) in the real-world clinical practice, and especially their dose-

response relationships. To address these knowledge gaps, HF was included as an outcome of

interest in the context of an overall cardiovascular and gastrointestinal risk evaluation of individual

NSAIDs within the Safety of Non-Steroidal Anti-Inflammatory (SOS) Project, a multinational

project funded by the European Commission under the 7th

Framework Program [http://www.sos-

nsaids-project.org]. A large, common-protocol, multinational, nested case-control study based on

electronic healthcare databases (DBs) from 4 European countries was carried out.

Methods

Data sources

This study is based on five electronic health DBs from four European countries, i.e. the Netherlands

(PHARMO), Italy (SISR and OSSIFF), Germany (GePaRD) and United Kingdom (THIN). Overall,

these databases covered a total of over 30 million individuals with different data-availability

periods between 1999 and 2010. Table 1 summarizes the main characteristics of these databases.

Further details are reported elsewhere [19].

Harmonization and data processing

Since these databases differed in several aspects, including the type of collected information (i.e.

healthcare utilization, claims, and primary care data), the classification systems used for disease and

medication coding (9th

or 10th

revisions of the International Classification of Diseases [ICD;

http://www.who.int/classifications/icd/en/] dictionaries for diseases coding; Anatomic-Therapeutic-

Chemical [ATC; http://www.whocc.no/atc/] or British National Formulary/Multilex [BNF/Multilex;

http://www.fdbhealth.com/multilex-drug-terminology/] dictionaries for medicines coding), a

harmonization of variable and outcome definitions was performed. Specifically, the Unified

Medical Language System (UMLS) [20] was used to harmonize the clinical diagnosis and

Page 5 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

5

conditions across databases, while the ATC classification system was used to harmonize

medications.

Anonymized data were extracted locally and processed with the software Jerboa©, developed by

Erasmus MC, providing individual-level datasets in a harmonized data format. These datasets were

securely transferred to the SOS Project data-warehouse, hosted by the University of Milano-

Bicocca, to be analysed in a centralized and secure way [21].

Cohort selection and follow-up

A cohort of NSAID users was selected from all DBs. Adults (≥18 years) who received at least one

NSAID prescription or dispensation (ATC: M01A) during 2000-2010 were considered eligible to

enter the cohort. The first recorded prescription or dispensation in the same period was defined as

the index prescription. Subjects were excluded if: i) they did not have at least one year of

uninterrupted observation prior to the index prescription date; ii) they received one or more NSAID

prescriptions or dispensations within in the year preceding the index prescription date (i.e. prevalent

NSAIDs users); iii) they received a diagnosis of malignant cancer (with the exception of non-

melanoma skin cancers); or iv) they were hospitalized for HF in the year before the index

prescription.

Each cohort member accumulated person-years of follow-up from the date of index prescription

(cohort entry) until the earliest among the outcome onset date (i.e. admission date for the first

hospitalization for HF) or censoring date (i.e. end of registration in the DB due to death or

emigration, diagnosis of malignancies other than non-melanoma skin cancers, or end of the DB-

specific data availability).

Cases and controls

A case-control study was nested into the cohort of new users of NSAIDs. The endpoint of interest

Page 6 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

6

was the first hospitalization for HF (i.e. with HF as the main cause or reason for the hospitalization)

identified during follow-up. HF is a heterogeneous clinical syndrome which involves several

pathophysiological mechanisms which, along with factors triggering circulatory decompensation,

may give a diversity of clinical manifestations which often receive a delayed diagnosis. Therefore,

our endpoint definition did not include i) diagnostic codes for clinical HF in the outpatient setting or

ii) secondary hospital discharge codes for HF (which likely represent HF manifestations occurring

during hospitalizations for other causes).

Consequently, cases were all cohort members hospitalized for HF for their first time during follow-

up, identified either from primary hospital discharge diagnoses (PHARMO, SISR, OSSIFF,

GePaRD) or codes registered by the general practitioner (THIN). The admission date of the first

hospitalization for HF identified during follow-up was defined as the index date. Codes used to

identify HF cases in each DB are reported in the supplementary material (Table S1).

Each case was matched to up to 100 controls randomly selected by incidence density sampling from

the cohort. Matching was performed within each DB on the basis of gender, age at cohort entry (± 1

year), and date of index prescription (± 28 days).

Exposure to NSAIDs

All NSAID prescriptions received by cohort members during follow-up were identified. Use of

individual NSAIDs was assessed according to the time elapsed between the last corresponding

prescription or dispensation and the index date. Specifically, cohort members were classified into

the following categories of current, recent, and past NSAID users. Current users were all patients

covered by a NSAID prescription or dispensation at the index date or the preceding 14 days. The

remaining patients were defined as recent users if they were covered during the 15-183 days before

the index date or as past users otherwise. The number of days’ supply of each NSAID prescription

or dispensation was computed from the corresponding dispensed amount of active principle and the

Page 7 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

7

daily dose. This was taken equal to the prescribed daily dose, if available (i.e. only in PHARMO

and THIN), or to a daily dose equivalent to 1 DDD (i.e. in all other DBs).

Covariates

Several covariates were assessed for each cohort member if available in the corresponding DB,

including: i) prior history of outpatient or secondary hospital diagnoses of HF, co-morbidities (e.g.

other CV diseases), and lifestyle or clinical characteristics (e.g. smoking status, alcohol abuse and

obesity), assessed in the 12 months before cohort entry; ii) concomitant use of specific drugs (e.g.

nitrates, diuretics, other drugs for CV diseases), assessed in the 90 days before the index date. Co-

morbidities were assessed using hospital discharge diagnoses in any position (PHARMO, GePaRD,

SISR, OSSIFF), outpatient clinical diagnoses (in GePaRD) and clinical electronic GP records (in

THIN), as well use of specific drugs as proxies for certain diseases and conditions not well recorded

in the databases. Table 2 below reports all collected covariates.

Statistical analysis

Individual-level data from all DBs were firstly gathered into a single pooled dataset using a

common data model and analysed by means of a multivariable conditional logistic regression model

[22,23]. The Odds Ratio (ORs), with 95% confidence intervals (CIs), measuring the association

between NSAIDs use (i.e. current use of individual NSAIDs or recent use of any NSAID in

comparison with past use of any NSAID) and the risk of hospitalization for HF were estimated.

Among the abovementioned covariates, only those available in all DBs entered the model.

Because DBs differed with respect to covered populations as well as type and level of detail of

available covariates, we evaluated the robustness of the pooled estimates using a meta-analytic

approach by means of the following procedure. We firstly separately fitted a conditional logistic

regression model to estimate the effect of each individual NSAID within every DB. To avoid

computational issues (i.e. model convergence failure due to sparse data), only NSAIDs with at least

Page 8 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

8

five exposed cases were considered in the model. The covariates available for all DBs were always

forced to enter the model. Other covariates were included provided they: 1) reached at least a 5%

prevalence among controls, 2) resulted significantly associated with the outcome in a univariate

analysis, and 3) were selected from a backward selection procedure (p-value for removal: p>0.10).

A random-effects meta-analytic model [24] was subsequently used to synthetize DB-specific ORs

into a summarized odds ratio (ORs), with 95% CI, for current use of each individual NSAID

(admitted that a point estimate was available from at least two DBs) contrasted with past use of any

NSAID. Heterogeneity between DB-specific ORs was assessed by means of Cochran’s Q and

Higgins’ I2 statistics [25].

Dose-response analysis

A dose-response analysis was performed to assess how the risk of HF hospitalization associated

with current use of individual NSAIDs varied along the considered prescribed daily dose categories.

As Italian and German DBs did not report data on prescribed daily doses, this analysis was

performed by pooling individual-level data gathered from DBs from the Netherlands (PHARMO)

and the United Kingdom (THIN). Patients for which the information on the prescribed daily dose

was not available in the corresponding DB of membership were not retained in the resulting pooled

dataset.

In this dataset, the prescribed daily dose for each individual NSAID was expressed in defined daily

dose equivalents and categorized as low (≤ 0.8 defined daily doses), normal (from 0.9 to 1.2), high

(from 1.3 to 1.9) or very high dose (≥ 2 defined daily dose equivalents) with respect to the

corresponding defined daily dose [http://www.whocc.no/ddd/]. To avoid computational issues in the

conditional logistic regression models, only NSAIDs for which all the considered categories

included at least one case of HF were considered in the analysis. Tests for trends in ORs were

performed.

Page 9 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

9

All statistical analyses were implemented using the SAS© software (v9.3; SAS Institute Inc. Cary,

NC, USA). All tests were two-sided and considered statistically significant for p-values p<0.05.

Results

Subjects

The flow-chart describing the attrition of eligible NSAIDs users after applying the exclusion criteria

is reported in the Supplementary Figure S1. Among the almost 10 million new users of NSAIDs

identified across all DBs, 7,680,181 met the inclusion criteria and constituted the study cohort.

Cohort members accumulated 24,555,063 person-years of follow-up and generated 92,163 cases of

hospitalized HF (incident rate of 37.5 HF events every 10,000 person-years). Cases were matched

to 8,246,403 controls.

As shown in Table 2, mean age (SD) was 77 (11) and 76 (10) respectively among cases and

controls. About 45% of both cases and controls were men. Compared to controls, cases had more

comorbidities (mainly cardiovascular disease, such as acute myocardial infarction, other ischemic

heart diseases, atrial fibrillation and flutter and valvular disease and endocarditis) and more often

received concomitant drug therapies (e.g., anticoagulants, cardiac glycosides, nitrates and Cyp2C9

inhibitors). About 9.1% of cases and 2.5% of controls already had a history of clinical diagnoses of

HF, either recorded as outpatient or a secondary hospital diagnoses in the year prior to starting

therapy with NSAIDs (cohort entry). The distribution of all considered covariates among cases and

controls in each DB is reported in Supplementary Table S2.

Use of NSAIDs and HF risk

Respectively 14,930 (16.20%) and 126,099 (13.76%) of all cases and matched controls identified

from all DBs were current users of an individual NSAID, while 28,647 (31.08%) and 279,711

(30.52%) of cases and controls were recent users of any NSAIDs. Figure 1 reports the distribution

Page 10 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

10

of current use of individual NSAIDs among all cases and controls. Among cases, the most

frequently used tNSAIDs were diclofenac (3.50%), nimesulide (2.95%), and ibuprofen (2.18%),

while the most frequently used COXIBs were celecoxib (1.36%), rofecoxib (1.32%), and etoricoxib

(0.91%) in the current period. The DB-specific distributions of NSAIDs use status among cases and

controls are reported in the Supplementary Table S3.

In the overall pooled analysis, subjects currently using any NSAID had a 20% higher HF risk than

past users (OR: 1.20; 95% CI: 1.17, 1.22). Conversely, there was no evidence that recent use of any

NSAID involved differences in HF risk with respect to past use (1.00; 0.99, 1.02). As shown in

Figure 1, a statistically significant higher risk of HF was observed in association with current use of

any of the following nine NSAIDs (compared with past use of any NSAIDs): ketorolac (1.83; 1.66,

2.02), etoricoxib (1.51; 1.41, 1.62), indomethacin (1.36; 1.28, 1.44), rofecoxib (1.36; 1.28, 1.44),

piroxicam (1.27; 1.19, 1.35), diclofenac (1.19; 1.15, 1.24), ibuprofen (1.18; 1.12, 1.23), nimesulide

(1.18; 1.14, 1.23), and naproxen (1.16; 1.07, 1.27). Other less frequently used NSAIDs (e.g.

sulindac) were also found to be associated with an increased risk of HF, although results were not

statistically significant.

In addition to the above nine individual NSAIDs, current use of nabumetone was also found

associated with a higher HF risk in the meta-analytic approach (Figure 2). Although between-DB

heterogeneity was relevant (I2 > 70%) for 4 NSAIDs (etoricoxib, diclofenac, ibuprofen, and

naproxen), meta-analytic OR estimates were generally consistent with the corresponding ones

obtained from the analysis of pooled individual-level data. The DB-specific relative risk estimates

for each individual NSAIDs are reported in Supplementary Table S4.

Dose-response relationship

A total of 20 (0.24%) cases and 855 (0.13%) controls from PHARMO and 753 (4.25%) cases and

61,777 (4.30%) controls from THIN were excluded as they were not associated with prescribed

Page 11 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

11

daily dose information. All remaining cases and controls from PHARMO and THIN entered the

analysis on the relationship between the NSAIDs dose currently prescribed and the HF risk. This

relationship was investigated for etoricoxib, indomethacin, rofecoxib, piroxicam, diclofenac,

ibuprofen, naproxen, celecoxib, and diclofenac combinations as these were the only NSAIDs for

which at least one expose case was identified in each dose category. The distribution of cases and

controls according the different dose categories of current NSAIDs use varied across individual

NSAIDs and is reported in Supplementary Table S5. As shown in Figure 3, current users of very-

high doses of diclofenac, etoricoxib, indomethacin, piroxicam, and rofecoxib had a more than two-

fold higher risk of HF than past users. For ibuprofen, the OR associated with current use at very

high-dose was also compatible with an increased risk despite the wide confidence intervals. Finally,

there was little evidence that celecoxib increased the risk of hospitalized HF even at very high doses

compared with past use of any NSAIDs.

Discussion

Our study, which is based on “real-world” data on almost 10 million NSAIDs users from four

European countries, provides evidence that current use of both selective and traditional non-

selective COX inhibitors are associated with an increase the risk of HF and that this association

varies between individual NSAIDs according to the used dose.

Specifically, we found a statistically significant increase in the risk of hospitalization for HF in

association with current use of several tNSAIDs (such as diclofenac, ibuprofen, indomethacin,

ketorolac, naproxen, nimesulide and piroxicam, and possibly nabumetone) and two COXIBs

(etoricoxib and rofecoxib). We also observed the increase in risk is dose dependent for most of the

individual NSAIDs that we could evaluate based on the available data. This is not surprising since

the overall effects of the inhibition of prostaglandin synthesis mediated by NSAIDs are to raise

peripheral systemic resistance and reduce renal perfusion in susceptible people triggering HF, and

Page 12 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

12

this inhibition increases with dose [17, 26]. Additionally, indomethacin and etoricoxib also

appeared to increase the risk of HF hospitalization even if used at medium doses.

All these findings extend those of the meta-analysis of randomized trials published by the CNT

collaboration showing that the risk of hospitalization due to HF was roughly doubled by all studied

NSAID regimens compared with placebo (COXIBs 2.28, 95% CI 1.62, 3.20; diclofenac 1.85,

1.17, 2.94; ibuprofen 2.49, 1.19, 5.20; naproxen 1.87, 1.10, 3.16) [11]. Similarly, a meta-analysis

of six trials comparing different NSAIDs showed no differences in HF risk between tNSAIDs and

COXIBs [16]. The estimates provided by the few published observational studies on the NSAID-HF

association were compatible with an increased HF risk associated with naproxen, ibuprofen,

ketoprofen, piroxicam, indomethacin, and rofecoxib, but not for celecoxib [15,17 27-30].

Our study did not find evidence that celocoxib, the most widely prescribed selective COX-2

inhibitor, increases the risk of HF hospitalization as used in the current clinical practice. This seems

to be the case also when celecoxib was used at the highest doses, although a moderate increase in

HF risk cannot be excluded. These results are nevertheless compatible with the body of evidence

supporting the relative CV safety of celecoxib for the treatment of arthritis compared to all other

COXIBs if used at low-medium doses [31].

Taken together, our findings support the hypothesis that both selective and non-selective COX-2

inhibitors increase the risk HF, but that the magnitude of this effect varies between individual drugs

according to the used dose [13]. The effect of individuals NSAIDs may in fact depend upon a

complex interaction of pharmacological properties, including duration and extent of platelet

inhibition, extent of blood pressure increase and properties possibly unique to the molecule [31].

The present study is based on data from very large and unselected populations from four European

countries. Data were derived from 5 healthcare databases which collectively represented a precious

source of information to investigate the safety of a large number of individual NSAIDs. In fact, the

Page 13 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

13

heterogeneity of these sources should be considered a strength of the SOS project, since it allowed

to compare the risk of HF associated with a large number of individuals NSAIDs as used in

different populations and healthcare systems.

Some limitations should however be acknowledged. First, our study may not have captured all

NSAID exposure, since many of these drugs (e.g. ibuprofen) are also available without prescription

(i.e. over the counter, OTC) in all four countries. The implication of this is that patients classified as

non-current users of NSAIDs in this study may actually have been current users of OTC NSAIDs.

Since such misclassification would tend to bias estimates toward the null [32,33], our findings may

understate the actual association between use of individual NSAIDs and HF risk.

Second, HF diagnoses could not be validated in all DBs for privacy and logistical concerns.

Validation was only possible in the Italian OSSIFF database, in which a positive predictive value

for cases of hospitalization due to HF was 80% (95% CI: 66%, 90%). Although this result does not

necessarily generalize to the other DBs, other investigations based on healthcare databases found

that specific hospital discharge HF diagnosis codes considered in this study (i.e. ICD-9 codes 428.*

or ICD-10 codes I50.*) had high PPV [33]. In addition, given their high impact [28], HF diagnoses

are likely to be recorded with high accuracy in all participating DBs (including the GP database

from the UK), either for reimbursement purposes or because of their importance for routine clinical

care. In fact, the incidence of almost 37.5 HF cases every 10,000 person-years observed in this

study does not substantially differ from rates reported by available population-based investigations

[34]. Nevertheless, even admitting some outcome misclassification [35], this is expected to be non-

differential, i.e. independent of current use of NSAIDs, leading to a bias towards underestimation of

the observed effect [36].

Third, a portion of patients registered in the PHARMO and THIN databases had to be excluded

from the dose-response analysis because they lacked the necessary prescribed daily dose

information. Although this might have led to some bias [37], the number of excluded subjects was

Page 14 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

14

low. Hence, it seems implausible that their exclusion could have had a significant impact on the

results.

Fourth, the impact of heterogeneity in baseline patients’ characteristics must be considered when

interpreting our findings. Indeed, some individual NSAIDs more frequently used for different acute

or chronic indications could have resulted in different patterns of use, as well as in different types of

populations of users. To address this possibility, our pooled estimates were adjusted for several

patients’ baseline demographic, therapeutic, and clinical characteristics (including osteoarthritis,

rheumatoid arthritis and inflammatory polyarthritis) measured in all the included data sources. In

addition, estimates did not substantially change in the random-effects meta-analytic approach,

where DB-specific estimates were adjusted for all baseline covariates available in the considered

data source, further protecting our conclusions. Moreover, we examined the HF risk associated with

the use of individual NSAIDs among patients with prior history of outpatient or secondary hospital

diagnoses of HF (i.e. patients with a contraindication for NSAIDs use who also have an important

risk factor for acute clinical HF). The corresponding relative risk estimates were similar to those

observed in the overall analysis (Supplementary Table S6), although power was low for some

NSAIDs studied in this subgroups analysis.

Lastly, residual confounding must also be considered. This is related to the fact that some diseases

that modify both the risk of HF and the probability of current NSAID use may have not been fully

accounted for in this study. To protect against this possibility, all our estimates were adjusted for

concomitant (i.e. in the current period) use of specific drugs (e.g. nitrates, diuretics, other drugs for

CV diseases) as a proxy of patients’ current health status. Still, residual confounding cannot be

excluded. For example, gout is potentially an uncontrolled confounder of the association between

current use of NSAIDs and HF risk in this study. This is because i) gout is an independent risk

factor for HF [38] and ii) NSAIDs are the first pharmacological choice for treating acute gout

episodes [39]. However, assuming that gout has a 1% prevalence in our source population and that

Page 15 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

15

it increases HF risk by 1.74-fold [40], we estimated [41] that acute gout episodes should increase

the odds of being treated with naproxen (the NSAID with the weakest statistically significant

association with HF among those investigated) in the current rather than the past period by 33-fold

(an implausibly high amount) to fully explain the observed naproxen-HF association. These

considerations further strengthen our conclusions.

In conclusion, our study offers further evidence that the most frequently used traditional NSAIDs

and selective COX-2 inhibitors are associated with an increased risk of hospitalization for heart

failure. Moreover, the risk appears to vary between molecules and according to the dose. For the

less frequently used NSAIDs we were not able to exclude a risk of low-moderate magnitude due to

the limited numbers of exposed cases identified in this study. Since any potential increased risk may

result in a considerable public health impact, particularly among the elderly, the risk effect

estimates provided by this study may help inform both clinical practices and regulatory activities.

Conflicts of Interests

Giovanni Corrao collaborated with the advisory boards of Novartis and Roche and participated in

projects funded by GSK. Huub Straatman; Ron Herings: are employees of the PHARMO Institute.

This independent research institute performs financially supported studies for government and

related healthcare authorities and several pharmaceutical companies. Bianca Kollhorst and Tania

Schink are working in departments that occasionally perform studies for the pharmaceutical

companies. These include Bayer-Schering, Celgene, GSK, Mundipharma, Novartis, Purdue, Sanofi-

Aventis, Sanofi-Pasteur, Stada, and Takeda. Edeltraut Garbe is running a department that

occasionally performs studies for pharmaceutical industries. These companies include Bayer,

Celgene, GSK, Mundipharma, Novartis, Sanofi, Sanofi Pasteur MSD, and STADA. EG has been a

consultant to Bayer, Nycomed, Teva, GSK, Schwabe and Novartis. SOS was not (co)-funded by

any of these companies. Silvia Lucchi and Marco Villa, as employees of the Local Health Authority

Page 16 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

16

of Cremona, have perfomed research studies sponsored by pharmaceutical companies (Pfizer Italia,

GlaxoSmithKline and Novartis V&D) unrelated to this study. Cristina Varas-Lorenzo, as employee

of RTI Health Solutions, worked on projects funded by pharmaceutical companies including

manufacturers of treatments for pain and inflammation and also participates in advisory boards

funded by pharmaceutical companies. Martijn J. Schuemie, has since completion of this research

accepted a full-time position at Janssen R&D. Vera Valkhoff, as employee of Erasmus MC, has

conducted research for AstraZeneca. Miriam Sturkenboom is head of a unit that conducts some

research for pharmaceutical companies: Pfizer, Novartis, Lilly and Altana. The SOS Project was not

(co)-funded by any of these companies. All other authors have no conflicts of interest to declare.

Acknowledgements and Funding

The research leading to the results of this study has received funding from the European

Community’s Seventh Framework Programme under grant agreement number 223495 - the SOS

project. We thank all members of the SOS project consortium for their collaborative efforts

(http://www.sos-nsaids-project.org/).

References

[1] FitzGerald GA, Patrono C. The coxibs, selective inhibitors of cyclooxygenase-2. New Engl J

Med 2001; 345:433–42

[2] Fries JF. Selective cyclooxygenase inhibition: promise for future NSAID therapy? Scand J

Rheumatol Suppl 1996; 102:23-8

[3] Bombardier C, Laine L, Reicin A, et al.; VIGOR Study Group. Comparison of upper

gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. N Engl J

Med 2000; 343:1520-8

Page 17 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

17

[4] Amer M, Bead VR, Bathon J, et al. Use of nonsteroidal anti-inflammatory drugs in patients with

cardiovascular disease. A cautionary tale. Cardiology Rev 2010; 18:204-12

[5] Bresalier RS, Sandler RS, Quan H. Cardiovascular events associated with rofecoxib in a

colorectal adenoma chemoprevention trial. New Engl J Med 2005; 352:1092-102

[6] Solomon SD, McMurray JJ, Pfeffer MA. Cardiovascular risk associated with celecoxib in a

clinical trial for colorectal adenoma prevention. New Engl J Med 2005; 352:1071-80

[7] Kearney PM, Baigent C, Godwin J, et al. Do selective cyclo-oxygenase-2 inhibitors and

traditional non-steroidal anti-inflammatory drugs increase the risk of atherothrombosis? Meta-

analysis of randomised trials. BMJ 2006; 332:1302-8

[8] García Rodríguez LA, Tacconelli S, Patrignani P. Role of dose potency in the prediction of risk

of myocardial infarction associated with nonsteroidal anti-inflammatory drugs in the general

population. J Am Coll Cardiol 2008; 52:1628-36

[9] McGettigan P, Henry D. Cardiovascular risk with non-steroidal anti-inflammatory drugs:

systematic review of population-based controlled observational studies. PLoS Med 2011;

8:e1001098

[10] Trelle S, Reichenbach S, Wandel S, et al. Cardiovascular safety of non-steroidal anti-

inflammatory drugs: network meta-analysis. BMJ 2011; 342:c7086

[11] Coxib and traditional NSAID Trialists' (CNT) Collaboration.

Vascular and upper

gastrointestinal effects of non-steroidal anti-inflammatory drugs: meta-analyses of individual

participant data from randomised trials. Lancet 2013; 382:769-79

[12] Fabule J, Adebajo A. Comparative evaluation of cardiovascular outcomes in patients with

osteoarthritis and rheumatoid arthritis on recommended doses of nonsteroidal anti-inflammatory

drugs. Ther Adv Musculoskelet Dis 2014; 6:111-30

Page 18 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

18

[13] Patrono C, Baigent C. Nonsteroidal anti-inflammatory drugs and the heart. Circulation 2014;

129:907-16

[14] Bleumink GS, Feenstra J, Sturkenboom MC, et al. Nonsteroidal anti-inflammatory drugs and

heart failure. Drugs 2003; 63:525-34

[15] Huerta C, Varas-Lorenzo C, Castellsague J, García Rodríguez LA. Non-steroidal anti-

inflammatory drugs and risk of first hospital admission for heart failure in the general population.

Heart 2006; 92:1610-5

[16] Scott PA, Kingsley GH, Scott DL. Non-steroidal anti-inflammatory drugs and cardiac failure:

meta-analysis of observational and randomized controlled trials. Eur J Heart Fail 2008; 10:1102-7

[17] Garcia Rodríguez LA, Hernández-Díaz S. Nonsteroidal antiinflammatory drugs as a trigger of

clinical heart failure. Epidemiology 2003; 14:240-6

[18] The Task Force for the Diagnosis and Treatment of Acute and Chronic Heart Failure 2012 of

the European Society of Cardiology. ESC Guidelines for the diagnosis and treatment of acute and

chronic heart failure 2012. European Heart Journal 2012; 33:1787–847.

[19] Valkhoff VE, Schade R, 't Jong GW, et al. Safety of non-steroidal snti-inflammatory drugs

(SOS) project. Population-based analysis of non-steroidal anti-inflammatory drug use among

children in four European countries in the SOS project: what size of data platforms and which study

designs do we need to assess safety issues? BMC Pediatr 2013; 13:192

[20] Avillach P, Mougin F, Joubert M, et al. A semantic approach for the homogeneous

identification of events in eight patient databases: a contribution to the European EU-ADR project.

Studies in health technology and informatics 2009; 150:190-4

Page 19 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

19

[21] Coloma PM, Schuemie MJ, Trifiro G, et al. Combining electronic healthcare databases in

Europe to allow for large-scale drug safety monitoring: the EU-ADR Project. Pharmacoepidemiol

DS 2011; 20:1-11

[22] Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd edition. John Wiley & Sons,

2000

[23] Cooper H, Patall EA. The relative benefits of meta-analysis conducted with individual

participant data versus aggregated data. Psychol Methods 2009; 14:165–76

[24] DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled Clinical Trials 1986;

7:177-88

[25] Higgins JPT, Thompson SG. Quantifying heterogeneity in a meta-analysis. Statist Med 2002;

21:1539-58

[26] Grosser T, Yu Y, Fitzgerald GA. Emotion recollected in tranquility: lessons learned from the

COX Back M, Yin L, Ingelsson E. Cyclooxygenase-2 inhibitors and cardiovascular risk in a nation-

wide cohort study after the withdrawal of rofecoxib. Eur Heart J 2012; 33:1928-33

[27] Bernatsky S, Hudson M, Suissa S. Anti-rheumatic drug use and risk of hospitalization for

congestive heart failure in rheumatoid arthritis. Rheumatology 2005; 44:677-80

[28] Mamdani M, Juurlink DN, Lee DS, et al. Cyclo-oxygenase-2 inhibitors versus non-selective

non-steroidal anti-inflammatory drugs and congestive heart failure outcomes in elderly patients: a

population-based cohort study. Lancet 2004; 363:1751-6

[29] Mangoni AA, Woodman RJ, Gaganis P, et al. Use of non-steroidal anti-inflammatory drugs

and risk of incident myocardial infarction and heart failure, and all-cause mortality in the Australian

veteran community. Br J Clin Pharmacol 2010; 69:689-700

Page 20 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

20

[30] McGettigan P, Han P, Jones L, et al. Selective COX-2 inhibitors, NSAIDs and congestive heart

failure: differences between new and recurrent cases. Br J Clin Pharmacol 2008 Jun;65(6):927-34.-2

saga. Annu Rev Med 2010; 61:17–33

[31] Howes LG. Selective COX-2 inhibitors, NSAIDs and cardiovascular events – is celocoxib the

safest choice? Ther Clin Risk Manag 2007; 3:831-45

[32] Yood MU, Campbell UB, Rothman KJ, et al. Using prescription claims data for drugs

available over-the-counter (OTC). Pharmacoepidemiol Drug Saf 2007; 16:961-8.

[33] Delaney JA, Biggs ML, Kronmal RA, et al. Demographic, medical, and behavioral

characteristics associated with over the counter non-steroidal anti-inflammatory drug use in a

population-based cohort: results from the Multi-Ethnic Study of Atherosclerosis.

Pharmacoepidemiol Drug Saf 2011; 20:83-9.

[34] Saczynski JS, Andrade SE, Harrold LR, et al. A systematic review of validated methods for

identifying heart failure using administrative data. Pharmacoepidemiol Drug Saf 2012; 21 Suppl 1:

129-140

[35] Corrao G, Ghirardi A, Ibrahim B, et al. Burden of new hospitalization for heart failure: a

population-based investigation from Italy. Eur J Heart Fail 2014; 16:729-36

[36] Romano PS, Mark DH. Bias in the coding of hospital discharge data and its implications for

quality assessment. Med Care 1994; 32:81-90

[37] Copeland KT, Checkoway H, McMichael AJ, et al. Bias due to misclassification in the

estimation of relative risk. Am J Epidemiol 1977; 105:488-95

[38] Altman DG, Bland JM. Missing data. BMJ 2007; 334:424.

Page 21 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

21

[39] Krishnan E. Gout and the risk for incident hearth failure and systolic dysfunction. BMJ Open

2012; 2:e000282

[40] Kim KY, Schumacher HR, Hunsche E, et al. A literature review of the epidemiology and

treatment of acute gout. Clin Ther 2003; 25:1593-617

[41] Smith EUR, Torné-Diaz C, Perez-Ruin F, et al. Epidemiology of gout: an update. Best Pract

Res Clin Rheumatol 2010; 24:811-27

[42] Schneeweiss S. Sensitivity analysis and external adjustment for unmeasured confounders in

epidemiologic database studies of therapeutics. Pharmacoepidemiol Drug Saf 2006; 15: 291–303

Page 22 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

22

Tables

Table 1. Databases considered as data sources in this study among those participating to the Safety

of Non-Steroidal Anti-Inflammatory (SOS) Project.

Country Database* Type

Size of the

covered

population

Covered

period

Diagnoses

coding Drugs coding

The

Netherlands

PHARMO (PHARMO Institute for

Drug Outcomes

Research)

Record

linkage

system

2.2 million 1999-2008 ICD-9-CMa ATC

d

Italy

SISR**

(Sistema Informativo

Sanitario Regionale)

Healthcare

Utilization

DB


d

OSSIFF (Osservatorio

Interaziendale per la

Farmacoepidemiologia e

la Farmacoeconomia)

Healthcare

Utilization

DB


d

Germany

GePaRD (German

Pharmacoepidemiological

Research Database)

Claims DB 13.7 million 2004-2009 ICD-10-GMb ATC

d

United

Kingdom

THIN (The Health Improvement

Network)

General

Practice DB 4.8 million 1999-2010 READ v2

c BNF/Multilex

e

* Other databases participated in the SOS Project but did not contribute data to this study. See Reference 19.

** Because OSSIFF covers a subset of patients also covered by SISR, this database excluded the common subset of

patients to avoid overlap.

a International Classification of Diseases, 9th revision, clinical modification [http://www.who. int/classifications/icd/en]

b International Classification of Diseases, 10th revision, German modification

c READ clinical classification system [Chisholm J. The Read clinical classification. BMJ 1990;300:1092]

d Anatomical-Therapeutic-Chemical classification system [http://www.whocc.no/atc/]

e British National Formulary codes / Multifunctional Standardised Lexicon for European Community Languages codes

[http://www.fdbhealth.com/multilex-drug-terminology]

Page 23 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

23

Table 2. Clinical features and other selected characteristics of the 92,163 patients hospitalized for

heart failure and the 8,246,403 matched control patients included into the study. The Safety of Non-

Steroidal Anti-Inflammatory (SOS) Project.

Case patients Controls N 92,163 8,246,403

Men 41,652 (45.2%) 3,671,565 (44.5%)

Age at cohort entry: mean (SD) 77 (11) 76 (10)

Comorbidities and other characteristics*

Acute Myocardial Infarction 3,063 (3.3%) 81,222 (1.0%)

Alcohol Abuse 1,942 (2.1%) 128,871 (1.6%)

Asthma 1,031 (1.1%) 57,079 (0.7%)

Atrial Fibrillation and Flutter 4,606 (5.0%) 110,217 (1.3%)

Chronic Liver Disease 1,815 (2.0%) 98,762 (1.2%)

Chronic Respiratory Disease 16,190 (17.6%) 870,497 (10.6%)

Diabetes 17,888 (19.4%) 725,320 (8.8%)

Heart Failure 8,353 (9.1%) 209,125 (2.5%)

Hyperlipidemia 18,793 (20.4%) 1,160,532 (14.1%)

Hypertension 19,905 (21.6%) 1,515,002 (18.4%)

Iron Deficiency Anemia 2,159 (2.3%) 83,926 (1.0%)

Ischemic Heart Disease 8,406 (9.1%) 294,986 (3.6%)

Kidney Failure 1,445 (1.6%) 41,094 (0.5%)

Obesity 4,555 (4.9%) 181,104 (2.2%)

Osteoarthritis 6,916 (7.5%) 483,721 (5.9%)

Other Cardiovascular Disease 13,055 (14.2%) 463,797 (5.6%)

RA and Inflammatory Polyarthritis 736 (0.8%) 40,269 (0.5%)

Smoking 164 (0.2%) 8,155 (0.1%)

Stroke Confounder 1,869 (2.0%) 85,109 (1.0%)

Valvular Disease and Endocarditis 2,383 (2.6%) 70,646 (0.9%)

Concomitant use of other drugs**

ACE Inhibitor/AT-II Antagonists 38,834 (42.1%) 2,030,050 (24.6%)

Anticoagulants 17,589 (19.1%) 442,725 (5.4%)

Aspirin 31,658 (34.4%) 1,669,443 (20.2%)

Beta Blockers 22,506 (24.4%) 1,253,749 (15.2%)

Calcium Channel Blockers 28,911 (31.4%) 1,754,965 (21.3%)

Cardiac Glycosides 14,429 (15.7%) 342,042 (4.1%)

Cyp2C9 Inducers 38 (0.0%) 1,149 (0.0%)

Cyp2C9 Inhibitors 8,289 (9.0%) 174,253 (2.1%)

Diuretics 48,991 (53.2%) 1,536,700 (18.6%)

Glucocorticoids 8,636 (9.4%) 349,012 (4.2%)

Nitrates 24,029 (26.1%) 717,669 (8.7%)

Platelet Aggregation Inhibitor 9,105 (9.9%) 367,716 (4.5%)

Vasodilators 1,654 (1.8%) 44,916 (0.5%)

* Assessed in the 12 months before cohort entry. Based on inpatient diagnoses, outpatient diagnoses (GePaRD only), medical

history (THIN only), and drug prescriptions as proxies. ** Assessed in the 14 days preceding the index HF hospitalization.

Page 24 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

24

Legend of figures

Figure 1. Distribution of current use of individual NSAIDs among all cases and controls and forest

plot of the pooled associations between current use of individual NSAIDs and the risk of

hospitalization for heart failure, with past use of any NSAID as reference. Estimates were obtained

by pooling individual data from all available databases.

Footnote. Pooled odds ratio (OR) and 95% confidence intervals (95% CI) were estimated by fitting a conditional

logistic regression model after correcting for available covariates (see text).

Figure 2. Forest plot of the summarized associations between of current use of individual NSAIDs

and the risk of hospitalization for heart failure, compared to past use of any NSAID. Estimates were

obtained by summarizing DBs specific estimates by the random-effects meta-analytic approach.

Footnote. Database-specific odds ratios were summarized into a unique odds ratio (ORs) admitted that it was

considered from at least two databases. ORs, and 95% confidence intervals (95% CI), were estimated by using a

random-effects model. Heterogeneity between DB-specific ORs was assessed by means of Cochran’s Q (and

corresponding p-value) and Higgins’ I2 statistics (see text). The number of summarized databases (N) is reported in the

last column

Figure 3. Dose-response relationship between the currently prescribed dose of certain NSAIDs and

the risk of heart failure with respect to past use of any NSAID.

Footnote. Pooled data from the Netherlands (PHARMO) and United Kingdom (THIN) DBs were used for this analysis.

The dose currently prescribed dose of each NSAID was categorized as low (0.8 daily dose equivalents), medium (from

0.9 to 1.2), high (from 1.3 to 1.9) and very high dose (≥ 2 defined daily dose equivalents). Odds ratio (OR) and 95%

confidence intervals (95% CI) were estimated by fitting a conditional logistic regression model after correcting for

available covariates (see text).

Page 25 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

25

Supplementary material

Figure S1. Flow-chart of inclusion/exclusion criteria.

Table S1. Codes considered to identify HF patients by the included databases.

Table S2. Clinical features and other selected characteristics of case patients hospitalized for heart

failure and matched controls included into each database.

Table S3. Database-specific distributions of NSAIDs use status among case and control.

Table S4. Adjusted odds ratio (ORs), and 95% confidence intervals (95% CI), from case-control

analyses in each database.

Table S5. Distributions cases and controls between the different considered dose categories for

current use of selected individual NSAIDs in the pooled PHARMO and THIN databases.

Table S6. Adjusted odds ratio (ORs), and 95% confidence intervals (95% CI), for association

between current use of individual NSAIDs and the risk of hospitalization for heart failure, with past

use of any NSAID as reference, among patients with recorded outpatient or secondary hospital HF

diagnoses prior to NSAIDs therapy initiation. Estimates were obtained by pooling individual data

from all available databases.

Page 26 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

Figure 1

214x170mm (300 x 300 DPI)

Page 27 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

Figure 2

214x115mm (300 x 300 DPI)

Page 28 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

Figure 3

170x189mm (300 x 300 DPI)

Page 29 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

1

Table S1. Codes considered to identify HF patients by the included databases.

Coding system Database

(Country) Codes/term

ICD-9-CMa SISR (IT)

OSSIFF (IT)

PHARMO (NL)

398.91, 402.01, 402.11, 402.91, 404.01, 404.03,

404.11, 404.13, 404.91, 404.93, 428, 428.0, 428.1,

428.9

ICD-10-GMb GePaRD (GER) I11.0, I11.00, I11.01, I13.0, I13.00, I13.01, I13.2,

I13.20, I13.21, I50, I50.0, I50.00, I50.01, I50.1,

I50.11, I50.12, I50.13, I50.14, I50.19, I50.9

READc v2 THIN (UK) 14A6.00, G1yz100, G21z100, G232.00, G234.00,

G58..00, G58..11, G580.00, G580.11, G580.12,

G580.13, G580.14, G580000, G580100, G580200,

G580300, G581.00, G581.11, G581.12, G581.13,

G581000, G582.00, G58z.00, G58z.11, G58z.12

a International Classification of Diseases, 9th revision, clinical modification [http://www.who.

int/classifications/icd/en]

b International Classification of Diseases, 10th revision, German modification

c READ clinical classification system [Chisholm J. The Read clinical classification. BMJ 1990;300:1092]

Page 30 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

2

Table S2. Clinical features and other selected characteristics of case patients hospitalized for

heart failure and matched controls included into each database.

S2.1 THE NETHERLANDS - PHARMO (PHARMO Institute for Drug Outcomes

Research)

Case patients

(N=8,250)

Controls

(N=653,466)

Men 4,646 (60.46%) 426,158 (62.33%)



Acute Myocardial Infarction 73 (0.88%) 2130 (0.33%)

Alcohol Abuse 16 (0.19%) 500 (0.08%)

Asthma 9 (0.11%) 199 (0.03%)

Atrial Fibrillation and Flutter 127 (1.54%) 2803 (0.43%)

Chronic Liver Disease 8 (0.10%) 334 (0.05%)

Chronic Respiratory Disease 1949 (23.62%) 86416 (13.22%)

Diabetes 1611 (19.53%) 52329 (8.01%)

Heart Failure 134 (1.62%) 2131 (0.33%)

Hyperlipidemia 1843 (22.34%) 91339 (13.98%)

Hypertension 702 (8.51%) 36948 (5.65%)

Iron Deficiency Anemia 361 (4.38%) 12444 (1.90%)

Ischemic Heart Disease 231 (2.80%) 7376 (1.13%)

Kidney Failure n.a.

Obesity 16 (0.19%) 948 (0.15%)

Osteoarthritis 157 (1.90%) 10831 (1.66%)

Other Cardiovascular Disease 519 (6.29%) 11422 (1.75%)

Rheumatoid Arthritis and Inflammatory Polyarthritis 18 (0.22%) 506 (0.08%)

Smoking 9 (0.11%) 279 (0.04%)

Stroke Confounder 26 (0.32%) 625 (0.10%)

Valvular Disease and Endocarditis 24 (0.29%) 403 (0.06%)


ACE Inhibitor/AT-II Antagonists 3861 (50.24%) 192401 (28.14%)

Anticoagulants 1909 (24.84%) 44910 (6.57%)

Aspirin 1466 (19.08%) 56614 (8.28%)

Beta Blockers 3854 (50.15%) 201893 (29.53%)

Calcium Channel Blockers 2255 (29.34%) 116932 (17.10%)

Cardiac Glycosides 1193 (15.52%) 23416 (3.43%)

Cyp2C9 Inducers 6 (0.07%) 113 (0.02%)

Cyp2C9 Inhibitors 496 (6.01%) 8132 (1.24%)

Diuretics 4256 (55.38%) 113224 (16.56%)

Glucorticoids 801 (10.42%) 33115 (4.84%)

Nitrates 1163 (15.13%) 33499 (4.90%)

Platelet Aggregation Inhibitor 629 (8.18%) 18744 (2.74%)

Vasodilators 520 (6.77%) 15409 (2.25%)

* Assessed in the 12 months preceding cohort entry. ** Assessed in the 14 days preceding the index HF hospitalization. n.a.: not available

Page 31 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

3

S2.2 ITALY – SISR (Sistema Informativo Sanitario Regionale)

Case patients

(N=34,773)

Controls

(3,312,454)

Men 14,140 (40.66%) 1,338,754 (40.42%)



Acute Myocardial Infarction 956 (2.75%) 22,591 (0.68%)

Alcohol Abuse 73 (0.21%) 2,704 (0.08%)

Asthma 79 (0.23%) 3,499 (0.11%)


Chronic Liver Disease 378 (1.09%) 15,486 (0.47%)


Diabetes 8,139 (23.41%) 371,582 (11.22%)

Heart Failure 3,833 (11.02%) 91,385 (2.76%)

Hyperlipidemia 7,334 (21.09%) 526,980 (15.91%)

Hypertension 10,215 (29.38%) 854,865 (25.81%)

Iron Deficiency Anemia n.a.


Kidney Failure 133 (0.38%) 3,460 (0.10%)

Obesity 438 (1.26%) 9,758 (0.29%)

Osteoarthritis 646 (1.86%) 48,086 (1.45%)


Rheumatoid Arthritis and Inflammatory Polyarthritis 157 (0.45%) 7,153 (0.22%)

Smoking n.a.

Stroke Confounder 414 (1.19%) 21,469 (0.65%)

Valvular Disease and Endocarditis 510 (1.47%) 10,053 (0.30%)


ACE Inhibitor/AT-II Antagonists 16,991 (48.86%) 974,943 (29.43%)

Anticoagulants 6,999 (20.13%) 189,656 (5.73%)

Aspirin 12,786 (36.77%) 759,562 (22.93%)

Beta Blockers 7,627 (21.93%) 435,123 (13.14%)

Calcium Channel Blockers 12,617 (36.28%) 861,069 (25.99%)


Cyp2C9 Inducers n.a.

Cyp2C9 Inhibitors 3,999 (11.50%) 90,114 (2.72%)

Diuretics 17,507 (50.35%) 556,149 (16.79%)

Glucorticoids 3,062 (8.81%) 130,582 (3.94%)

Nitrates 10,567 (30.39%) 338,230 (10.21%)


Vasodilators n.a.


Page 32 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

4

S2.3 ITALY – OSSIFF (Osservatorio Interaziendale per la Farmacoepidemiologia e

la Farmacoeconomia)

Case patients

(N=23,753)

Controls

(N=2,159,548)

Men 9,924 (41.78%) 884,886 (40.98%)




Alcohol Abuse 58 (0.24%) 2,334 (0.11%)

Asthma 61 (0.26%) 2,396 (0.11%)

Atrial Fibrillation and Flutter 859 (3.62%) 19,174 (0.89%)

Chronic Liver Disease 233 (0.98%) 9,287 (0.43%)


Diabetes 3,899 (16.41%) 153,790 (7.12%)

Heart Failure 1,504 (6.33%) 36,585 (1.69%)

Hyperlipidemia 2,864 (12.06%) 193,743 (8.97%)

Hypertension 5,241 (22.06%) 408,379 (18.91%)

Iron Deficiency Anemia 625 (2.63%) 28,750 (1.33%)


Kidney Failure 46 (0.19%) 913 (0.04%)

Obesity 130 (0.55%) 2120 (0.10%)

Osteoarthritis 521 (2.19%) 31,274 (1.45%)



Smoking n.a.





Anticoagulants 4,520 (19.03%) 121,319 (5.62%)

Aspirin 7,275 (30.63%) 398,196 (18.44%)

Beta Blockers 3,223 (13.57%) 183,367 (8.49%)



Cyp2C9 Inducers 18 (0.08%) 673 (0.03%)

Cyp2C9 Inhibitors 2,567 (10.81%) 56,793 (2.63%)

Diuretics 11,539 (48.58%) 339,377 (15.72%)

Glucorticoids 1,895 (7.98%) 77,641 (3.60%)

Nitrates 6,850 (28.84%) 205,666 (9.52%)


Vasodilators n.a.


Page 33 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

5

S2.4 GERMANY – GePaRD (German Pharmacoepidemiological Research

Database)

Case patients

(N=7,685)

Controls

(N=683,663)

Men 3,937 (47.72%) 303,716 (46.48%)



Acute Myocardial Infarction 1,239 (16.12%) 37,070 (5.42%)

Alcohol Abuse 229 (2.98%) 7,702 (1.13%)

Asthma 610 (7.94%) 36,198 (5.29%)


Chronic Liver Disease 1,166 (15.17%) 72,751 (10.64%)


Diabetes 1,944 (25.30%) 71,867 (10.51%)

Heart Failure 2,767 (36.01%) 77,676 (11.36%)

Hyperlipidemia 2,506 (32.61%) 149,564 (21.88%)

Hypertension 2,439 (31.74%) 150,496 (22.01%)



Kidney Failure 1,176 (15.30%) 34,817 (5.09%)

Obesity 1,737 (22.60%) 76,194 (11.14%)

Osteoarthritis 2,233 (29.06%) 187,447 (27.42%)



Smoking n.a.


Valvular Disease and Endocarditis 1,500 (19.52%) 52,696 (7.71%)



Anticoagulants 1,909 (24.84%) 44,910 (6.57%)

Aspirin 1,466 (19.08%) 56,614 (8.28%)

Beta Blockers 3,854 (50.15%) 201,893 (29.53%)



Cyp2C9 Inducers 5 (0.07%) 64 (0.01%)

Cyp2C9 Inhibitors 496 (6.01%) 8,132 (1.24%)

Diuretics 4,256 (55.38%) 113,224 (16.56%)

Glucorticoids 801 (10.42%) 33,115 (4.84%)

Nitrates 1,163 (15.13%) 33,499 (4.90%)

Platelet Aggregation Inhibitor 629 (8.18%) 18,744 (2.74%)

Vasodilators 520 (6.77%) 15,409 (2.25%)


Page 34 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

6

S2.5 UNITED KINGDOM – THIN (The Health Improvement Network)

Case patients

(N=17,702)

Controls

(N=1,437.272)

Men 9,005 (50.87%) 718,051 (49.96%)




Alcohol Abuse 1,566 (8.85%) 115,631 (8.05%)

Asthma 272 (1.54%) 14,787 (1.03%)

Atrial Fibrillation and Flutter 289 (1.63%) 7,444 (0.52%)

Chronic Liver Disease 30 (0.17%) 904 (0.06%)


Diabetes 2,295 (12.96%) 75,752 (5.27%)

Heart Failure 115 (0.65%) 1,348 (0.09%)

Hyperlipidemia 4,246 (23.99%) 198,906 (13.84%)

Hypertension 1,308 (7.39%) 64,314 (4.47%)



Kidney Failure 87 (0.49%) 1,849 (0.13%)

Obesity 2,234 (12.62%) 92,084 (6.41%)

Osteoarthritis 3,359 (18.98%) 206,083 (14.34%)

Other Cardiovascular Disease 738 (4.17%) 18,969 (1.32%)


Smoking 155 (0.88%) 7,876 (0.55%)



Concomitant use of other drugs*


Anticoagulants 2,252 (12.72%) 41,931 (2.92%)

Aspirin 8,665 (48.95%) 398,457 (27.72%)

Beta Blockers 3,948 (22.30%) 231,473 (16.11%)



Cyp2C9 Inducers 9 (0.05%) 299 (0.02%)

Cyp2C9 Inhibitors 731 (4.13%) 11082 (0.77%)

Diuretics 11,433 (64.59%) 414,726 (28.86%)

Glucorticoids 2,077 (11.73%) 74,559 (5.19%)

Nitrates 4,286 (24.21%) 106,775 (7.43%)


Vasodilators 614 (3.47%) 14,098 (0.98%)


Page 35 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

7

Table S3. Database-specific distributions of NSAIDs use status among case and control.

S3.1 THE NETHERLANDS - PHARMO (PHARMO Institute for Drug Outcomes

Research)

NSAIDs exposure Case patients

(N=8,250)

Controls

(N=653,466)

Current Use of:

Aceclofenac 3 (0.04%) 261 (0.04%)

Acemetacin n.a.

Azapropazone 2 (0.02%) 58 (0.01%)

Celecoxib 47 (0.57%) 2,825 (0.43%)

Dexibuprofen 2 (0.02%) 90 (0.01%)

Dexketoprofen 0 (0.00%) 7 (0.00%)

Diclofenac 380 (4.61%) 21,456 (3.28%)

Diclofenac, combinations 144 (1.75%) 8,273 (1.27%)

Etodolac n.a.

Etoricoxib 69 (0.84%) 2,363 (0.36%)

Fenbufen n.a.

Fenoprofen n.a.

Fentiazac n.a.

Flurbiprofen 0 (0.00%) 87 (0.01%)

Ibuprofen 171 (2.07%) 10,700 (1.64%)

Ibuprofen, combinations n.a.

Indometacin 32 (0.39%) 992 (0.15%)

Ketoprofen 6 (0.07%) 441 (0.07%)

Ketoprofen, combinations n.a.

Ketorolac n.a.

Lonazolac n.a.

Lornoxicam n.a.

Lumiracoxib n.a.

Meclofenamic acid n.a.

Mefenamic acid n.a.

Meloxicam 93 (1.13%) 5,574 (0.85%)

Mofebutazone n.a.

Morniflumate n.a.

Nabumetone 17 (0.21%) 919 (0.14%)

Naproxen 137 (1.66%) 7,303 (1.12%)

Naproxen and esomeprazole n.a.

Niflumic acid n.a.

Nimesulide n.a.

Oxaprozin n.a.

Parecoxib n.a.

Phenylbutazone 1 (0.01%) 13 (0.00%)

Prioxicam 18 (0.22%) 873 (0.13%)

Proglumetacin n.a.

Rofecoxib 111 (1.35%) 5,234 (0.80%)

Sulindac 13 (0.16%) 331 (0.05%)

Tenoxicam 0 (0.00%) 15 (0.00%)

Tiaprofenic acid 2 (0.02%) 219 (0.03%)

Tolfenamic acid 0 (0.00%) 7 (0.00%)

Valdecoxib 0 (0.00%) 31 (0.00%)

Recent Use of any NSAID 1,725 (20.91%) 142,047 (21.74%)

n.a.: not available

Page 36 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

8

S3.2 ITALY – SISR (Sistema Informativo Sanitario Regionale)


(N=34,773)

Controls

(3,312,454)

Current Use of:

Aceclofenac 224 (0.64%) 21,005 (0.63%)

Acemetacin 0 (0.00%) 10 (0.00%)

Azapropazone n.a.

Celecoxib 532 (1.53%) 59,505 (1.80%)

Dexibuprofen 32 (0.09%) 2,294 (0.07%)

Dexketoprofen 0 (0.00%) 1 (0.00%)

Diclofenac 1,010 (2.90%) 74,937 (2.26%)


Etodolac n.a.

Etoricoxib 411 (1.18%) 30,116 (0.91%)

Fenbufen n.a.

Fenoprofen n.a.

Fentiazac 1 (0.00%) 47 (0.00%)

Flurbiprofen 13 (0.04%) 1224 (0.04%)

Ibuprofen 514 (1.48%) 31,834 (0.96%)

Ibuprofen, combinations n.a.

Indometacin 95 (0.27%) 5262 (0.16%)

Ketoprofen 483 (1.39%) 44,291 (1.34%)

Ketoprofen, combinations 0 (0.00%) 1 (0.00%)

Ketorolac 239 (0.69%) 9,993 (0.30%)

Lonazolac n.a.

Lornoxicam 34 (0.10%) 3,046 (0.09%)

Lumiracoxib n.a.

Meclofenamic acid 0 (0.00%) 13 (0.00%)

Mefenamic acid 0 (0.00%) 84 (0.00%)

Meloxicam 214 (0.62%) 21,162 (0.64%)

Mofebutazone n.a.

Morniflumate 1 (0.00%) 10 (0.00%)

Nabumetone 18 (0.05%) 1,967 (0.06%)

Naproxen 121 (0.35%) 11,748 (0.35%)


Niflumic acid 0 (0.00%) 1 (0.00%)

Nimesulide 1,770 (5.09%) 128,443 (3.88%)

Oxaprozin 16 (0.05%) 2,440 (0.07%)

Parecoxib n.a.

Phenylbutazone n.a.

Prioxicam 531 (1.53%) 41,483 (1.25%)

Proglumetacin 8 (0.02%) 873 (0.03%)

Rofecoxib 394 (1.13%) 27,740 (0.84%)

Sulindac 1 (0.00%) 132 (0.00%)

Tenoxicam 32 (0.09%) 2,907 (0.09%)


Tolfenamic acid n.a.

Valdecoxib 23 (0.07%) 1,787 (0.05%)

Recent Use of any NSAID 12,916 (37.14%) 1,213,352 (36.63%)

n.a.: not available

Page 37 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

9

S3.3 ITALY – OSSIFF (Osservatorio Interaziendale per la Farmacoepidemiologia e

la Farmacoeconomia)


(N=23,753)

Controls

(N=2,159,548)

Current Use of:

Aceclofenac 62 (0.26%) 6,757 (0.31%)

Acemetacin 0 (0.00%) 21 (0.00%)

Azapropazone n.a.

Celecoxib 374 (1.57%) 33,872 (1.57%)

Dexibuprofen 9 (0.04%) 876 (0.04%)

Dexketoprofen n.a

Diclofenac 592 (2.49%) 42,466 (1.97%)


Etodolac n.a.

Etoricoxib 194 (0.82%) 10,370 (0.48%)

Fenbufen n.a

Fenoprofen n.a

Fentiazac 0 (0.00%) 23 (0.00%)

Flurbiprofen 15 (0.06%) 1,117 (0.05%)

Ibuprofen 186 (0.78%) 11,194 (0.52%)

Ibuprofen, combinations 0 (0.00%) 1 (0.00%)

Indometacin 53 (0.22%) 2,942 (0.14%)

Ketoprofen 251 (1.06%) 21,191 (0.98%)

Ketoprofen, combinations n.a

Ketorolac 210 (0.88%) 7,462 (0.35%)

Lonazolac n.a

Lornoxicam 15 (0.06%) 1,200 (0.06%)

Lumiracoxib n.a

Meclofenamic acid n.a

Mefenamic acid 0 (0.00%) 23 (0.00%)

Meloxicam 121 (0.51%) 11,721 (0.54%)

Mofebutazone n.a

Morniflumate 0 (0.00%) 31 (0.00%)

Nabumetone 8 (0.03%) 1,173 (0.05%)

Naproxen 96 (0.40%) 7,413 (0.34%)

Naproxen and esomeprazole n.a

Niflumic acid 0 (0.00%) 38 (0.00%)

Nimesulide 947 (3.99%) 68,944 (3.19%)

Oxaprozin 13 (0.05%) 1207 (0.06%)

Parecoxib n.a

Phenylbutazone n.a

Prioxicam 376 (1.58%) 28,903 (1.34%)

Proglumetacin 7 (0.03%) 431 (0.02%)

Rofecoxib 412 (1.73%) 29,140 (1.35%)

Sulindac 2 (0.01%) 26 (0.00%)

Tenoxicam 18 (0.08%) 1,655 (0.08%)


Tolfenamic acid n.a

Valdecoxib 12 (0.05%) 617 (0.03%)


n.a.: not available

Page 38 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

10

S3.4 GERMANY – GePaRD (German Pharmacoepidemiological Research

Database)


(N=7,685)

Controls

(N=683,663)

Current Use of:

Aceclofenac 1 (0.01%) 262 (0.04%)

Acemetacin 14 (0.18%) 818 (0.12%)

Azapropazone n.a.

Celecoxib 20 (0.26%) 1,268 (0.19%)

Dexibuprofen 3 (0.04%) 328 (0.05%)

Dexketoprofen 6 (0.08%) 425 (0.06%)

Diclofenac 692 (9.00%) 48,954 (7.16%)

Diclofenac, combinations 13 (0.17%) 994 (0.15%)

Etodolac n.a.

Etoricoxib 80 (1.04%) 2,733 (0.40%)

Fenbufen n.a.

Fenoprofen n.a.

Fentiazac n.a.

Flurbiprofen n.a.

Ibuprofen 507 (6.60%) 28,337 (4.14%)

Ibuprofen, combinations

Indometacin 21 (0.27%) 996 (0.15%)

Ketoprofen 2 (0.03%) 206 (0.03%)

Ketoprofen, combinations n.a.

Ketorolac n.a.

Lonazolac 0 (0.00%) 5 (0.00%)

Lornoxicam 1 (0.01%) 78 (0.01%)

Lumiracoxib 0 (0.00%) 87 (0.01%)


Mefenamic acid n.a.

Meloxicam 25 (0.33%) 1,249 (0.18%)

Mofebutazone 0 (0.00%) 3 (0.00%)

Morniflumate n.a.

Nabumetone 1 (0.01%) 55 (0.01%)

Naproxen 19 (0.25%) 898 (0.13%)


Niflumic acid n.a.

Nimesulide n.a.

Oxaprozin n.a.

Parecoxib 0 (0.00%) 10 (0.00%)

Phenylbutazone 1 (0.01%) 113 (0.02%)

Prioxicam 25 (0.33%) 1,301 (0.19%)

Proglumetacin 1 (0.01%) 97 (0.01%)

Rofecoxib n.a.

Sulindac n.a.

Tenoxicam n.a.


Tolfenamic acid n.a.

Valdecoxib 0 (0.00%) 62 (0.01%)


n.a.: not available

Page 39 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

11

S3.5 UNITED KINGDOM – THIN (The Health Improvement Network)


(N=17,702)

Controls

(N=1,437,272)

Current Use of:

Aceclofenac 6 (0.03%) 473 (0.03%)

Acemetacin 2 (0.01%) 130 (0.01%)

Azapropazone 1 (0.01%) 213 (0.01%)

Celecoxib 280 (1.58%) 21,455 (1.49%)

Dexibuprofen 1 (0.01%) 80 (0.01%)

Dexketoprofen 2 (0.01%) 95 (0.01%)

Diclofenac 554 (3.13%) 53,979 (3.76%)


Etodolac 40 (0.23%) 3,578 (0.25%)

Etoricoxib 81 (0.46%) 4,457 (0.31%)

Fenbufen 3 (0.02%) 214 (0.01%)

Fenoprofen 0 (0.00%) 35 (0.00%)

Fentiazac n.a.

Flurbiprofen 2 (0.01%) 353 (0.02%)

Ibuprofen 634 (3.58%) 53,880 (3.75%)

Ibuprofen, combinations 1 (0.01%) 171 (0.01%)

Indometacin 66 (0.37%) 3364 (0.23%)

Ketoprofen 7 (0.04%) 821 (0.06%)

Ketoprofen, combinations 0 (0.00%) 4 (0.00%)

Ketorolac n.a.

Lonazolac n.a.

Lornoxicam 0 (0.00%) 31 (0.00%)

Lumiracoxib n.a.


Mefenamic acid 4 (0.02%) 802 (0.06%)

Meloxicam 176 (0.99%) 14,785 (1.03%)

Mofebutazone n.a.

Morniflumate n.a.

Nabumetone 22 (0.12%) 1,184 (0.08%)

Naproxen 217 (1.23%) 15,035 (1.05%)

Naproxen and esomeprazole 0 (0.00%) 1 (0.00%)

Niflumic acid n.a.

Nimesulide n.a.

Oxaprozin n.a.

Parecoxib n.a.

Phenylbutazone 0 (0.00%) 1 (0.00%)

Prioxicam 24 (0.14%) 1,862 (0.13%)

Proglumetacin n.a.

Rofecoxib 296 (1.67%) 16,816 (1.17%)

Sulindac 0 (0.00%) 150 (0.01%)

Tenoxicam 1 (0.01%) 139 (0.01%)


Tolfenamic acid 0 (0.00%) 21 (0.00%)

Valdecoxib 3 (0.02%) 304 (0.02%)


n.a.: not available

Page 40 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Only12

Table S4. Adjusted Odds Ratio (ORs), with 95% Confidence Intervals (95% CI), measuring the association between current use of individual

NSAIDs (compared with past use of any NSAID) and risk of a hospitalization for heart failure from the database-specific case-control analyses.

PHARMO (NL) SISR (IT) OSSIFF (IT) GEPARD (GER) THIN (UK)

OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI

Aceclofenac n.a. 1.10 (0.96, 1.25) 0.82 (0.64, 1.06) n.a. n.a.

Acemethacin n.a. n.a. n.a. 1.27 (0.73, 2.20) n.a.

Celecoxib 1.13 (0.84, 1.52) 0.90 (0.83, 0.98) 0.99 (0.89, 1.10) 1.20 (0.76, 1.90) 0.98 (0.87, 1.11)

Diclofenac 1.35 (1.21, 1.50) 1.27 (1.19, 1.35) 1.25 (1.14, 1.36) 1.33 (1.21, 1.45) 0.92 (0.84, 1.01)

Diclofenac, combination 1.34 (1.12, 1.59) 0.92 (0.72, 1.17) 0.94 (0.71, 1.26) 0.93 (0.53, 1.64) 0.93 (0.80, 1.08)

Etodolac n.a. n.a. n.a. n.a. 0.85 (0.62, 1.17)

Etoricoxib 2.23 (1.73, 2.88) 1.34 (1.21, 1.48) 1.62 (1.40, 1.88) 2.04 (1.61, 2.59) 1.41 (1.13, 1.77)

Ibuprofen 1.36 (1.16, 1.60) 1.20 (1.09, 1.31) 1.23 (1.06, 1.43) 1.46 (1.32, 1.61) 1.01 (0.93, 1.10)

Indomethacin 2.28 (1.58, 3.28) 1.40 (1.13, 1.72) 1.33 (1.01, 1.76) 1.67 (1.07, 2.62) 1.54 (1.20, 1.98)

Ketoprofen n.a. 1.05 (0.96, 1.15) 1.01 (0.89, 1.15) n.a. n.a.

Ketorolac n.a. 1.74 (1.52, 1.98) 1.99 (1.72, 2.30) n.a. n.a.

Meloxicam 1.15 (0.93, 1.42) 1.02 (0.89, 1.17) 1.00 (0.83, 1.20) 1.54 (1.02, 2.32) 0.92 (0.79, 1.07)

Nabumetone 1.46 (0.88, 2.40) n.a. n.a. n.a. 1.50 (0.97, 2.31)

Naproxen 1.39 (1.16, 1.66) 0.90 (0.75, 1.07) 1.15 (0.94, 1.41) 1.51 (0.94, 2.42) 1.23 (1.07, 1.42)

Nimesulide n.a. 1.20 (1.14, 1.26) 1.17 (1.09, 1.25) n.a. n.a.

Piroxicam 1.88 (1.16, 3.04) 1.27 (1.16, 1.39) 1.24 (1.12, 1.38) 1.70 (1.12, 2.57) 1.01 (0.67, 1.52)

Rofecoxib 1.61 (1.31, 1.96) 1.43 (1.29, 1.59) 1.26 (1.13, 1.39) n.a. 1.30 (1.15, 1.46)

Sulindac 1.89 (1.06, 3.38) n.a. n.a. n.a. n.a.

Recent Use of any

NSAID

0.98 (0.92, 1.04) 0.98 (0.96, 1.01) 0.98 (0.95, 1.01) 1.03 (0.97, 1.09) 1.09 (1.05, 1.14)

n.a.: not available

Page 41 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

13

Table S5. Distributions cases and controls between the different considered dose categories for

current use of selected individual NSAIDs in the pooled PHARMO and THIN databases. Only

patients for which the prescribed daily dose of current NSAIDs use was recorded in the

corresponding database of membership were retained in this analysis.


(N=25,179)

Controls

(N=2,028,106)

Celecoxib

Low 25 (9.3%) 1,459 (7.2%)

Medium 212 (78.8%) 16,735 (82.8%)

High 7 (2.6%) 403 (2.0%)

Very High 25 (9.3%) 1,609 (8.0%)

Diclofenac

Low 59 (8.1%) 3,541 (6.3%)

Medium 215 (29.3%) 16,904 (30.1%)

High 408 (55.7%) 34,060 (60.7%)

Very High 51 (7.0%) 1,613 (2.9%)

Diclofenac, combinations

Low 16 (5.7 %) 1,032 (5.2%)

Medium 131 (48.0%) 10,366 (51.8%)

High 122 (44.7%) 8,493 (42.4%)

Very High 4 (1.5%) 133 (0.7%)

Etoricoxib

Low 2 (1.7%) 165 (3.1%)

Medium 57 (47.5%) 2,646 (49.3%)

High 40 (33.3%) 1,898 (35.4%)

Very High 21 (17.5%) 655 (12.2%)

Ibuprofen

Low 107 (19.6%) 7,659 (17.9%)

Medium 383 (70.0%) 28,960 (67.7%)

High 52 (9.5%) 5,901 (13.8%)

Very High 5 (0.9%) 250 (0.6%)

Indometacin

Low 9 (13.4%) 471 (16.0%)

Medium 35 (52.2%) 1,462 (49.6%)

High 21 (31.3%) 954 (32.4%)

Very High 2 (3.0%) 62 (2.1%)

Naproxen

Low 11 (4.0%) 703 (4.2%)

Medium 59 (21.7%) 3,677 (21.9%)

High 47 (17.3%) 2,707 (16.1%)

Very High 155 (57.0%) 9,711 (57.8%)

Piroxicam

Low 4 (11.1%) 177 (8.1%)

Medium 28 (77.8%) 1,796 (82.6%)

High 1 (2.8%) 57 (2.6%)

Very High 3 (8.3%) 145 (6.7%)

Rofecoxib

Low 107 (33.8%) 5,963 (36.3%)

Medium 197 (62.2%) 9,977 (60.7%)

High 3 (1.0%) 106 (0.7%)

Very High 10 (3.2%) 381 (2.3%)


The dose currently prescribed dose of each NSAID was categorized as

low (0.8 daily dose equivalents), medium (from 0.9 to 1.2), high (from

1.3 to 1.9) and very high dose (≥ 2 defined daily dose equivalents).

Page 42 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

14

Table S6. Adjusted odds ratio (ORs), and 95% confidence intervals (95% CI), for association

between current use of individual NSAIDs and the risk of hospitalization for heart failure, with past

use of any NSAID as reference, among patients with recorded outpatient or secondary hospital HF

diagnoses prior to NSAIDs therapy initiation. Estimates were obtained by pooling individual data

from all available databases.


(N=8,353)

Controls

(N=185,761) OR 95% CI

Current use of:

Ketorolac 39 (0.47%) 501 (0.70%) 2.05 (1.24, 3.40)

Piroxicam 93 (1.11%) 1,607 (0.87%) 1.66 (1.21, 2.27)

Etoricoxib 86 (1.03%) 1,222 (0.66%) 1.66 (1.21, 2.26)

Rofecoxib 84 (1.01%) 1,339 (0.72%) 1.50 (1.06, 2.12)

Meloxicam 44 (0.53%) 871 (0.47%) 1.48 (1.00, 2.20)

Ibuprofen 303 (3.63%) 4,357 (2.35%) 1.28 (1.10, 1.50)

Diclofenac 401 (4.80%) 6,982 (3.76%) 1.18 (1.03, 1.36)

Indometacin 22 (0.26%) 316 (0.17%) 1.15 (0.65, 2.04)

Nimesulide 253 (3.03%) 4,735 (2.55%) 1.10 (0.91, 1.34)

Aceclofenac 28 (0.34%) 672 (0.36%) 1.08 (0.64, 1.83)

Diclofenac, comb. 22 (0.26%) 361 (0.19%) 1.01 (0.56,1.80)

Naproxen 20 (0.24%) 576 (0.31%) 0.97 (0.54, 1.73)

Ketoprofen 66 (0.79%) 1,494 (0.80%) 0.95 (0.67, 1.33)

Celecoxib 85 (1.01%) 2,127 (1.15%) 0.87 (0.63, 1.20)

Recent Use of any NSAID 2,964 (35.48%) 64,322 (34.63%) 0.95 (0.88, 1.02)

Page 43 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

Supplementary Figure 1

150x150mm (300 x 300 DPI)

Page 44 of 43


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

bmj · 2016-10-02 · unfortunately, reports of cardiovascular (cv) adverse reactions began to...

Documents